Fixed Camera Wiring Diagram

[kgranat]

Our original assembly guide had an incorrect wiring diagram for the camera. If you wired up your camera before April 29th 4:00 CST, please check your camera against the latest version of the assembly guide that now has the correct wiring diagram the white and green data lines need to be swapped from the incorrect graphic.

The incorrect wiring should not have caused any damage to your system and there should be extra heat shrink for you to make repairs, but if you have problems (like length problems)please let me know.



Attachment 5896

[jveejay]

Hey there,

Where can I find the specs on the camera? Also, is there any reason that we can/not use the raspi camera with the Pi?

[tician]

Where can I find the specs on the camera?

Is it not simply the cheap RobotGeek VGA-resolution USB webcam?

Also, is there any reason that we can/not use the raspi camera with the Pi?

As I recall, one reason given for not using the RPiCamera was difficulty in safely routing the flat-flex cable within the confines of the robot's shell. Also, a USB webcam provides a uniform sensor for all kits since the initial design was for Edison which has no camera interface like the RPi. Using the RPiCamera can also require a bit more effort to integrate into code than a standard USB camera with a V4L driver/interface.

[jwatte]

There exists a v4l driver for the RPi camera now, but it's not higher performance than the USB.
The benefit of the RPi camera comes when you use the RPi-only APIs to route it straight into the GPU, and use the GPU for H.264 encode, JPEG encode, image processing, display, etc.

[jveejay]

There exists a v4l driver for the RPi camera now, but it's not higher performance than the USB.
The benefit of the RPi camera comes when you use the RPi-only APIs to route it straight into the GPU, and use the GPU for H.264 encode, JPEG encode, image processing, display, etc.

I originally built the bot fully using the Edison, but then the camera seemed to be a big issue since the performance is rather weak with the Edison. You can't get enough FPS doing real-time vision processing. Therefore, I removed it and went to the Raspi.

You do make a good point that the GPU based compute is good for real-time processing, and that's the real reason that I went with the RPi camera. Otherwise, it is really hard to get the CPU to do much if the system is busy processing the video stream.

I do realize that using the RPi camera locks me down to the Pi. I can see why Trossen guys decided to go the obvious platform independent approach.

Are you guys doing anything with real-time face detection and responsiveness from the robot? I'm basically trying to do face detection based social development of the robot.

[jveejay]

By the way, I have not used the RobotGeek VGA USB webcam. Do you guys have prior experience with it? Could someone describe your experience? I haven't played with it enough, so it is possible that it can be used very effectively, but perhaps I just coded it badly?!

[jwatte]

Otherwise, it is really hard to get the CPU to do much if the system is busy processing the video stream.

WIth the Rpi 2, this is less of a problem, especially if you use a compiler that supports Neon and turn it on with compile options.

[jveejay]

WIth the Rpi 2, this is less of a problem, especially if you use a compiler that supports Neon and turn it on with compile options.

Sweet. Thanks for letting me know. I was counting on using the RPi 2 camera, so that I can leverage the SoC fully for fast processing of video. I had to get an extension cable because the 150mm default RPi camera doesn't reach the main connector. I am hoping the longer one I got will do the trick, can't wait to get vision up and running because I can't think of many things to do without vision!

[LloydF]

Is there any code for the camera's yet I mean even a test, to show that they still work?

[Piper]

We were able to get them to work for face identification with OpenCV. We don't have any demo code on us, but we can confirm that the hardware works.